DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/irclogs/ Connected with DDNet's IRC channel, Matrix room and GitHub repositories — IRC: #ddnet on Quakenet | Matrix: #ddnet-developer:matrix.org GitHub: https://github.com/ddnet

Between 2022-09-12 00:00:00Z and 2022-09-13 00:00:00Z

Ryozuki

https://chadnauseam.com/coding/gamedev/automated-testing-in-bevy/

Was actually searching a good game engine Yesterday and found bevy

BASED

there was another one, forgot the name

07:22

I think it started by Fox.... smth

07:22

it has an editor

07:25

Fyrox

I can't seem to settle on how to make the new global bans thing. I want to make something generic enough that we stop having to bolt on different ways of interacting with the outside from the gameserver

Inter server comms

I was originally thinking a new protocol, but it made me feel like I was re-inventing the wheel

Imagine a server sharing the entire team state to another

09:14

And switching

09:14

When a ddos happens

I'm currently considering HTTP + SSE. With HTTP/2 persistent connections it wouldn't be too bad when I eventually want to hook the other direction up allowing servers to communicate with the hub

Ryozuki

Imagine a server sharing the entire team state to another

this makes another kind of DoS possible. You could hog the entire inter-server bus by just ddosing the servers as teams are switched one to the other

Why not ipc (edited)

Sadly the state is large enough that this isn't feasible

Or grpc

Ryozuki

Why not ipc (edited)

I considered ipc too even directly shared memory, but with tiny amount of traffic on that bus I really didn't think it'd be worth it

Maybe just make the servers so a http request

09:17

On a json file

09:17

Every x mins

09:18

Most simple solution imho

On the leaf servers I'm thinking http over unix domain sockets, the spoke proxies and the hubs would communicate over http over tcp like normally

And then a post request to add a ban

Ryozuki

Most simple solution imho

Most simple but it's too slow, bans should go through immediately. And it also suffers from waste as the file is usually unchanged

09:19

I'm mostly unsure about the http part honestly

Http bloat

09:19

BASED

Should I bother with devising a proper protocol or is http good enough with the relatively small amount of traffic?

09:20

If it wasn't for http/2 persistent connections I think it'd be very wasteful given the server will want to communicate backwards too, but with persistent connections I think it's feasible enough

If you ask me whats most fun, a protocol, if you want to just get things done then http

09:20

I would go with fun

What is grpc btw?

09:22

I'm guessing it's rpc over some exotic protocol well suited for it?

gRPC is a modern open source high performance Remote Procedure Call (RPC) framework that can run in any environment.

09:23

it uses http2

09:23

and doesnt work on browsers

09:23

its for apps

09:23

also uses protobuf

09:23

https://grpc.io/

A high-performance, open source universal RPC framework

ah they decided abusing http2 persistent connections were good enough for them

09:23

who am I to challange that

Image attachment

09:24

yeah

Funnily enough this really looks like the final design I'm considering. So maybe I'm not that far off from a good solution

09:24

They just used protobuf instead of the json I'm thinking about

https://github.com/hyperium/tonic

GitHub - hyperium/tonic: A native gRPC client & server implementati...

A native gRPC client & server implementation with async/await support. - GitHub - hyperium/tonic: A native gRPC client & server implementation with async/await support.

09:24

for rust

09:25

protobuf is probs more perf

And I guess they do the streaming data better than the sse I'm thinking of

09:28

Seems they support json too. I might just use grpc instead of making it from scratch. Thanks, I was sure someone else solved this problem before

Would be cool to do this as a rust module @Learath2

09:33

BASED

09:33

And you have the lib i sent

09:33

Im sure in c++ u need to pull a big lib

09:33

greenthing

Ryozuki

Would be cool to do this as a rust module @Learath2

I think we need this quicker than @heinrich5991's rust stuff can be added to the client

09:35

I'm considering just using grpc-json though. With that I think I hope I can just use curl on gameservers

Sadge

I'm still in general pretty skeptical about generic async programming. It just feels so wrong to let some scheduler handle my threading

Just make ur async runtime 4head

The problem isnt that the runtimes arent good enough. The problem is that when it gets to the runtime a lot of information is already lost

09:58

Maybe a runtime with task tagging could be nice. E.g. tasks with same tags run on the same thread pinned on one core e.g.

Hmm

10:54

Why you want such fine grained control?

10:54

Maybe then dont use a async runtime

10:54

But an actor system

10:55

I think runtimes are p good at managing tasks

10:57

Well idk if an actor system can give more control

11:00

https://github.com/bastion-rs/bastion

GitHub - bastion-rs/bastion: Highly-available Distributed Fault-tol...

Highly-available Distributed Fault-tolerant Runtime - GitHub - bastion-rs/bastion: Highly-available Distributed Fault-tolerant Runtime

Ryozuki

Why you want such fine grained control?

Idk, sometimes you just know better than the runtime. The runtime can only know what is happening right now, you can also make predictions about the whole running duration of the program. Say you do a calculation that'll be necessary in a couple ms. If the result of that calculation and the task it's necessary for happens on different cores now you've lost locality which'll be slower

https://docs.rs/tokio/latest/tokio/task/fn.spawn_local.html

spawn_local in tokio::task - Rust

Spawns a !Send future on the local task set.

11:04

Maybe this helps?

Learath2

Idk, sometimes you just know better than the runtime. The runtime can only know what is happening right now, you can also make predictions about the whole running duration of the program. Say you do a calculation that'll be necessary in a couple ms. If the result of that calculation and the task it's necessary for happens on different cores now you've lost locality which'll be slower

I guess at some point runtimes have to do tradeoffs (edited)

11:11

The reality is that often you dont need such control imho

11:11

And i think tokio is rly perfomant rn

I guess the idea with runtimes is that you usually don't care for that much optimization

11:13

Which is true, unless you are doing some highly computational stuff you'll be bound by other things before you are bound by how efficiently you are using your cpu

I mean, im sure the tokio devs care

11:13

But i guess its more "general optimization" than a specific usecase one

Ryozuki

I mean, im sure the tokio devs care

Yeah, but they can only optimize for the generic case, it's hard to do general optimizations for a specific usecase

Yea

Especially when it can be insanely architecture dependent. Like the fact that on some architectures some cores have more FPUs and some cores share some amount of cache

If you use a modern linux, for example there is tokio-uring

11:16

ye

@Learath2 there is also this

12:38

https://docs.rs/tokio/latest/tokio/task/index.html#unconstrained

tokio::task - Rust

Asynchronous green-threads.

12:38

Image attachment

@deen could you try https://github.com/ddnet/ddnet/pull/5599#issuecomment-1242691646 again?

WIP Rust PoC by heinrich5991 · Pull Request #5599 · ddnet/ddnet

blurb

heinrich5991

@deen could you try https://github.com/ddnet/ddnet/pull/5599#issuecomment-1242691646 again?

done

https://aeon.co/essays/ten-questions-about-the-hard-limits-of-human-intelligence https://www.sdfo.org/gj/stories/flowersforalgernon.pdf

Ten questions about the hard limits of human intelligence | Aeon Es...

Science and mathematics may never fully capture the physical universe. Are there hard limits to human intelligence?

16:03

good reads

@noby found smth better than your obfuscation

19:29

Image attachment

2

left-to-right arabic smh xd

pmcrts

xd

when arabic binary

I need help with this

Image attachment

read pinned message in #bugs

3D Tee Clones still long way to go. I just thought it was cool (in active development with godot) (edited)

20:59

21:00

Easter Egg Guess the map ;D

sone

3D Tee Clones still long way to go. I just thought it was cool (in active development with godot) (edited)

godot dev pog

21:01

do you use premade assets (classes and such) to make the movement like this?

Nope xD

all manual, damn

21:02

anywhere to see the code if you allow?

sure lemme upload it to github in 10-15mins

here you go! https://github.com/osnesone/TeeClones

GitHub - osnesone/TeeClones: TeeClones is inspired by Teeworlds fng...

TeeClones is inspired by Teeworlds fng/ctf/ddrace and CloneArmies - GitHub - osnesone/TeeClones: TeeClones is inspired by Teeworlds fng/ctf/ddrace and CloneArmies

sone

Click to see attachment 🖼️

I tried to do something similar, but it turned out that the control of the camera in 3d is too complicated to do basic movements like hammer hit or drag

I'll clean up the code tommor

21:17

Yes I remember you. It was pretty cool ngl but I am going in another direction. Not so much ddnet/ddrace style as fng/ctf style

1

Ryozuki

https://aeon.co/essays/ten-questions-about-the-hard-limits-of-human-intelligence https://www.sdfo.org/gj/stories/flowersforalgernon.pdf

the second looks a bit annoying to read due to spelling mistakes. still worth it despite this?

Exported 115 message(s)